Cellular Wave Computing in Nanoscale via Million Processor Chips
A bifurcation is emerging in computer science and engineering due to the sudden emergence of many-core or even kilo-processor chips on the market. Due to the physical limitations, in CMOS technologies below 65 nm, a drastic power dissipation limit, a major signal propagation speed and distance limit, and a distributed character of the circuit elements are forcing new architectures. As a result, locality, the local connectedness becomes a prevailing property, the cellular, i.e., mainly locally connected processor arrays are becoming the norm, and the cellular wave dynamics can produce unique and practical effects.
In this new world, new principles are needed and new design methodologies. Luckily, the 15 years of research and development in cellular wave computing and CNN technology, we have aquired skills that help establishing some principles and techniques that might lead toward a new computer science and technology in designing mega-processor systems from kilo-processor chips.
In this chapter, we review the architectural development from standard CNN dynamics to the Cellular Wave Computer, showing several practical implementations, introduce the basic concepts of the Virtual Cellular Machine, present a new kind of implementation combining spatial-temporal algorithms with physics, give some architectural principles for non-CMOS implementations, and comment on biological relevance.
KeywordsImage Flow Black Pixel Disjunctive Normal Form Processor Array Natural Noise
The supports of the Office of Naval Research, the Future and Emerging Technology program of the EU, the Computer and Automation Research Institute of the Hungarian Academy of Sciences, the Hungarian National Research Fund (OTKA), the Pázmány P. Catholic University, Budapest, the University of California at Berkeley, and the University of Notre Dame are gratefully acknowledged.
- Chua LO (1999) A paradigm for complexity. World Scientific, New York, SingaporeGoogle Scholar
- de-Souza SX, Suykens JAK, Vandewalle J (2006) Learning of spatiotemporal behavior in cellular neural networks. Int J Circuit Theor Appl 34:127–140Google Scholar
- Halfhill TR (2007) Faster than a blink. Microprocessor, www.MPRonline.com, 2/12/07, 2007
- ITRS (2007) International technology roadmap for semiconductors 2003, 2005, 2007Google Scholar
- Kék L, Karacs K, Zarándy Á, Roska T (2007) CNN template and subroutine library for cellular wave computing. Report DNS -1 – 2007, Computer and Automation Research Institute of the Hungarian Academy of Sciences, BudapestGoogle Scholar
- Kunz R, Tetzlaff R, Wolf D (2000) Brain electrical activity in epilepsy characterization of the spatio-temporal dynamics with cellular neural networks based on a correlation dimension analysis. IEEE Int Symp Circuits Syst (ISCAS 00)Google Scholar
- Roska B, Werblin FS (2001) Vertical interactions across ten parallel, stacked representations in the mammalian retina. Nature 410:583–587 (see also in Scientific American, April, 2007)Google Scholar
- Roska T (2007a) Cellular wave computers for nano-tera-scale technology – beyond boolean, spatial-temporal logic in million processor devices. Electron Lett 43:427–429 (Insight Letter)Google Scholar
- Szatmári I (2006) Object comparison using PDE-based wave metric on cellular neural networks. ibid, vol. 34, pp. 359–382, 2006.Google Scholar
- Turing A (1952) The chemical basis of morphogenesis. Phil Trans R Soc Lond 237B:37–72Google Scholar
- Von Neumann J (1987) Papers of John von Neumann on computing and computer theory. In Aspray W, Burks A (eds) Section IV: Theory of natural and artificial automata. The MIT Press and Tomash Publications, Los Angeles/San FranciscoGoogle Scholar